Fully automated futures trading

[FCT]

Brilliant and congrats! Does that include the stocks or just futures trading?

On the holding period: think that’s not really informative if rolls are included. Given that you trade long-term signals, we’re probably just looking at your weighted average futures roll cycle. Even a long-and-hold futures trader, who never changes position, would have a 30d hp simply because she has to roll futures.

 

[globalarbtrader]

Brilliant and congrats! Does that include the stocks or just futures trading?

On the holding period: think that’s not really informative if rolls are included. Given that you trade long-term signals, we’re probably just looking at your weighted average futures roll cycle. Even a long-and-hold futures trader, who never changes position, would have a 30d hp simply because she has to roll futures.

Includes stocks but both components are up.

Your comments about rolls are correct.

GAT

 

[Elder]

I assume I execute at the next days closing price, plus slippage.

For the EWMAC multiple there are 3 approaches you can take:

- use real data
- generate random data with trends plus noise http://qoppac.blogspot.co.uk/2015/11/using-random-data.html
- Apparently there is a closed form solution for the optimal given certain assumptions about the underlying price process, which a very clever guy once showed me. I couldn't reproduce it here for legal reasons as it's not in the public domain (plus I forgot).

Anyway, the right multiple seems to somewhere between 2 and 6 depending on which approach you take. And the performance surface is relatively flat between these values. So it doesn't make a lot of difference. If for example you ran 4,8; 4,12; 4,16; 4, 20; and 4,24 all at the same time; well the correlations between these things are very high ...

Since I like to use fast moving average of lookback 4,8,16 and so on (you need to double it to keep the correlation of adjacent EWMAC the same); using 4 times has a certain neatness to it.

GAT

Hi GAT, I am trying to work through and make sure my backtesting logic is sound. I note the industry practice, and your preference, is to use the 'current' day's signal on the next day's price for backtests, which I understand is linked to avoiding some sort of data snooping. However, if the system is set up to trade live based on the live signal during a trading day (as your system is) then presumably the closing price (and the signal associated with it) later in the trading day can just be considered random noise around the live price and live signal which you used to trade (and both are perfectly synchronised in this scenario). In which case, I would think it might be reasonable to apply the current day's forecast against the current day's closing price as opposed to the next day's closing price (for which the live price at which you traded is a proxy). Would you disagree?

 

[globalarbtrader]

Hi GAT, I am trying to work through and make sure my backtesting logic is sound. I note the industry practice, and your preference, is to use the 'current' day's signal on the next day's price for backtests, which I understand is linked to avoiding some sort of data snooping. However, if the system is set up to trade live based on the live signal during a trading day (as your system is) then presumably the closing price (and the signal associated with it) later in the trading day can just be considered random noise around the live price and live signal which you used to trade (and both are perfectly synchronised in this scenario). In which case, I would think it might be reasonable to apply the current day's forecast against the current day's closing price as opposed to the next day's closing price (for which the live price at which you traded is a proxy). Would you disagree?

Yes the backtesting logic is just to be conservative. In practice you'd take last nights close, calculate your forecast and then trade that during the subsequent trading day.

GAT

 

[Elder]

Thanks and congrats on your latest performance!

 

[isotope1]

I know this is off topic, but I just wanted to see if anyone here has any experience of writing automated market making strategies?

 

[truetype]

As expected, the London megatrendies reporting a big October. All AHL's programs now up on the year, as is Winton's.

 

[traider]

As expected, the London megatrendies reporting a big October. All AHL's programs now up on the year, as is Winton's.

What's causing this? Is it the US tax reform?

 

[truetype]

In a sense. Back of the envelope, the trend crowd made 15% long equities this year, and lost about the same thrashing around in other sectors.

 

[Spectre2007]

Most macro markets have been sideways trading in broad envelopes. If you didn't take profits at broad ATR zones, you essentially got chopped up. But trendies aren't expected to take profits, expected to let things ride. The only ride really was equities.